智慧型馬達旋轉故障診斷系統設計

Design and Implementation of Intelligent Diagnosis System for Motor Rotary Faults

王智豪(Zhi-Hao Wang)

指導教授 ： 曾傳蘆

國立臺北科技大學/電資學院/電機工程系所/碩士(2007年)

摘要

本論文主要以工廠馬達監測為出發點，提出一個整合型監測系統。前端的感測設備可放置於定點，再透過無線網路或是網路線的傳送可將所測得的資料送至後端的檢測系統。由於所測得的信號含有高頻雜訊，因此本論文使用適應性濾波器濾除量測信號中的雜訊，使所設計之故障診斷系統可以產生較好的鑑別率及提高鑑別速度。
本論文所設計之故障診斷系統包含振動訊號產生器、馬達實測振動資料庫、適應性濾波器以及智慧型故障診斷等功能模組。由於馬達轉動且即將發生轉動故障時，其頻域特徵會在基本頻率的二倍、三倍甚至是數倍的地方出現故障特徵。不同的故障會產生不同的頻率特徵。因此，本論文利用此特徵研製振動訊號產生器藉以訓練類神經故障診斷系統。資料庫中則儲存研究人員實際量測馬達運轉時所產生的振動訊號，以利後續分析。適應性濾波器模組則可以藉由學習的方式學習雜訊特徵，濾除量測訊號中的雜訊。濾除後的訊號則輸入訓練後之類神經故障診斷系統，判別其故障類別。除此之外，本系統提供適當的人機介面，使工程師可以藉由頻域波形清楚得知馬達運轉情形以及在運轉中的頻率。
最後，本論文利用MATLAB軟體撰寫馬達故障診斷系統。馬達振動資料庫中所儲存之振動資料來源為台北馬偕醫院、台北三軍總醫院以及台北市公所的冷卻水馬達的實測資料。由模擬與實測結果可知，本論文所提出之系統架構的確具有可行性與實用性。

關鍵字

故障檢測 ； 馬達監控 ； 類神經鑑別 ； 馬達故障診斷

並列摘要

This thesis proposes an integrated diagnosis system for motors in workplace. The front-end sensing modules are mounted on certain place of motor and the measured signals are transmitted by way of wireless network or internet to the back-end diagnosis system. Because the measured signal contains the high-frequency noise, in thesis, the adaptive filter is used to filter out the noise. Consequently, the designed diagnosis system can perform the better classification results and improve the speed of classification.
The proposed diagnosis system is composed of the motor vibration signal generator, motor vibration database, adaptive filter and intelligent fault diagnosis function modules. When the motor operates and rotary fault occurs, special patterns of characteristic frequencies may appear at double or triple, even several times of fundamental frequency. Because different faults generate different frequency characteristics, this research utilizes this fact to develop a motor vibration signal generator, of which the output signal simulates the vibration signal as motor operates. This generator can be used to train the neural-network based diagnosis system. The database stores the motor vibration signals measured in the factory or workplace. These signals facilitate the follow-up analysis. The adaptive filter equipped with learning capability can automatically adjust the filter gain according the noise characteristics. The filtered signals are taken as the input of the trained fault diagnosis system to classify the fault pattern. In addition, this integrated system offers suitable man-machine interfaces which provide the spectrum display window. From the display, the engineer can easily extract the useful information of motor operation and find the frequency pattern during the operation.
Finally, this thesis utilizes the MATLAB software to establish the motor diagnosis system. The signals stored in motor vibration database are obtained by measuring the motors of cooling water systems of Taipei Tri-Service General Hospital, Taipei Mackery Memorial Hospital, and the Xinyi district office of Taipei city government. From the results of simulation and implementation, it is seen that the proposed structure of motor diagnosis system is feasible and practicable.

並列關鍵字

Fault Detection ； Neural Network ； fault Diagnosis of Motor System ； Motor Monitoring.

參考文獻

[39] 趙安明，馬達故障診斷之模糊類神經網路，碩士，中原大學機械工程學系，中壢市，2004。

[1] B. Li and M. Y. Chow, “Neural-network-based motor rolling bearing fault diagnosis,” IEEE Transactions on Industrial Electronics, vol. 47, 2000, pp. 1060-1069.

[3] S. A. McInerny and Y. Dai, “Basic vibration signal processing for bearing fault detection,” IEEE Transactions on Education, vol. 46, February 2003, pp. 149-156.

[4] R. Szczesny, P. Kurzynski, H. Piqueb and K. Hwan, “Knowledge-base system approach to power electronic system fault diagnosis,” IEEE International Symposium on Industrial Electronics, vol. 2, 1996, pp. 1005-1010.

[7] S. Nandi and H. A. Roliyat, “Condition monitoring and fault diagnosis of electrical machine – a review,” IEEE Industry Applications Conference, vol. 1, 1999, pp. 197-204.